The present invention relates to methods of fabricating micromechanical resonators using Silicon-on-insulator (SOI) wafers, and in particular to methods of fabricating micromechanical resonators using single crystal silicon wafer material.
As disclosed by Yoon, et al. in U.S. Pat. No. 6,490,147, “High-Q Micromechanical Device and Method of Tuning Same,” Dec. 3, 2002, which is incorporated herein by reference, and also by Nguyen, et al. in U.S. Pat. No. 6,249,073, “Device Including a Micromechanical Resonator Having an Operating Frequency and Method of Extending Same,” Jun. 19, 2001, which is incorporated herein by reference, vibrating micromechanical resonators formed of polycrystalline silicon, commonly known as polysilicon, are well-known as miniaturized substitutes for crystals in a variety of high-Q oscillator and filtering applications. State of the art micromechanical resonator fabrication techniques utilize polycrystalline silicon for manufacturing micromechanical resonators devices produced by means of silicon sacrificial surface micromachining.
This state of the art method of manufacturing using sacrificial surface micromachining produces polysilicon resonators or other thin film resonators having intrinsic stresses, stress gradients, or both, that effect device performance. These intrinsic stresses and stress gradients are difficult to control during manufacturing. Yet, control of these intrinsic stresses and stress gradients is critical for devices for use in applications that require high repeatability and reproducibility.
This state of the art method of manufacturing using polysilicon also requires removal of a sacrificial layer of material using wet etching techniques that complicate the manufacturing process and generally results in low yield due to difficulty in removing sacrificial material in the small gap between the resonator and the lower electrode. This method of manufacturing using polycrystalline silicon also results in stiction, which further lowers yields.
For devices requiring very small capacitive air gaps, e.g., to compensate for manufacturing tolerances or for tuning or for use as tunable resonator cum filter, the removal of sacrificial layer becomes extremely difficult as liquid or even vapor etching techniques cannot easily penetrate underneath the resonators to create a free standing structures. The removal of etched sacrificial material to form the capacitive gap is a process as well as operation yield limiter for small air gaps on the order of 300 Angstrom or smaller.
Additionally, impurities present in polysilicon thin films degrade device performance and also result in lower resonator Q.
FIGS. 1 and 2 present conceptual and perspective view schematics, respectively, of a tunable capacitor 10 as taught by Yoon, et. al. in U.S. Pat. No. 6,490,147. The capacitor 10 has a bottom capacitor plate 12 fixed to a substrate 14, and a top capacitor plate 16 suspended above the bottom plate 12. The top capacitor plate 16 is also anchored to the substrate 14. Both plates 12 and 16 are constructed of copper (Cu) to minimize their total series resistance in an attempt to maximize the device quality factor, Q.
A dielectric slab 18 is suspended between the two plates 12 and 16 and anchored to the substrate 14 outside the two plates 12 and 16 via spring structures 20. This dielectric 18 is free to move by electrostatic displacement to alter either the overlap between it and the capacitor plates 12 and 16, or the fringing fields between them. In the former case, when a DC bias is applied between the two plates 12 and 16, the charges on the capacitor plates 12 and 16 exert an electrostatic force on the induced charges in the dielectric 18 to pull the dielectric 18 into the gap between the plates 12 and 16, as shown in FIG. 1. The “waffle” shape of the capacitor 10 shown in FIG. 2 is intended to minimize the travel distance, or the needed voltage, required for a given change in capacitance, and to provide etchant access paths during a step in the fabrication process for removing a thin sacrificial layer by etching.
FIGS. 3A-3E are side sectional views which illustrate one state of the art fabrication process for producing micromechanical resonators of the type depicted by the capacitor 10. The prior art process, as taught by Yoon, et. al. in U.S. Pat. No. 6,490,147, begins in FIG. 3A with the thermal growth of a 1 micron layer 30 of SiO.sub.2 to serve as an isolation or dielectric layer between the eventual metal structures and a silicon wafer or substrate 32. Next, the bottom capacitor plate 12 is formed by first evaporating 300 Angstroms/2000 Angstroms a Cr/Cu seed layer 34, then electroplating a 5 micron layer 36 of copper (Cu). A 3000 Angstrom layer 38 of nickel (Ni) is then electroplated above the Cu layer 36 to serve as a buffer layer to prevent Cu contamination of etch chambers during subsequent reactive ion etch (RIE) processes.
FIG. 3B illustrates a first 2000 Angstrom aluminum (Al) sacrificial layer 40 is evaporated and patterned to form vias through which a subsequent layer PECVD nitride dielectric film 42 adheres to the underlying Ni layer 38. The nitride film 42 is patterned via RIE to form the movable dielectric plate 18, then submerged under 0.9 micron of a second sacrificial Al film 44 that defines the spacing between the dielectric plate 18 and the eventual top metal plate 16, as shown in FIG. 3C. Due to the valley-like topography between the fingers of the etched dielectric, the deposition of the 0.9 micron layer 44 of Al actually results in only a 0.3 micron gap between the top plate 16 and the dielectric 18 when the two are engaged.
After etching vias through the Al layer 44 to define top plate anchors (shown in FIG. 3C), as shown in FIG. 3D, the top plate 16 is formed by first evaporating a thin Cr/Cu seed layer 46, then electroplating a Cu layer 48 through a defining photoresist mold 50 to a thickness sufficient to insure that the top plate 16 does not bend under applied actuation voltages. The PR and seed layer under the PR are removed. The two Al sacrificial layers 40 and 44 are selectively etched to release the dielectric 42 using a K.sub.3Fe(CN).sub.6/NaOH solution, which attacks Al, but leaves Cu and the nitride dielectric 42 intact, yielding the final cross-section shown in FIG. 3E. After release, sublimation or a critical point dryer is often used to dry the capacitor 10 in an attempt to prevent stiction.
Additionally, cleaning and removal of the sacrificial layer is extremely difficult for small gaps, and often requires use of a surfactant.
FIG. 4 illustrates a perspective view schematic of a free-free beam, flexural-mode, micromechanical device or resonator 52 and an electrical pick off scheme, as taught by Nguyen, et al. in U.S. Pat. No. 6,249,073. The device 52 includes a free-free micromechanical flexural resonator beam 54 supported at its flexural nodal points 56 by four torsional beams 58, each anchored to a substrate 59 by rigid contact anchors 60. A drive electrode 62 underneath the free-free resonator beam 54 allows electrostatic excitation via an applied AC voltage Vi, and output currents are detected directly off a DC-biased (via V.sub.P) resonator structure 64. The torsional support beams 58 are designed with quarter-wavelength dimensions, which effect an impedance transformation that isolates the free-free resonator beam 54 from the rigid anchors 60. Ideally, the free-free resonator beam 54 sees zero-impedance into its supports or beams 58, and thus, effectively operates as if levitated without any supports. As a result, anchor dissipation mechanisms normally found in previous clamped-clamped beam resonators are greatly suppressed, allowing much higher device Q. However, multiple drive electrodes may be utilized for push-pull excitation. The electrodes can also be used for sensing, frequency tuning and detection of the output.
Typically, a transducer capacitor gap spacing is entirely determined via a sacrificial surface micromachining process for removing a thin sacrificial oxide layer, and wet etching of the sacrificial layer for final release of the flexural resonator beam 54 to create the capacitor gap.
FIGS. 5A and 5B illustrate a transducer capacitor gap spacing, as taught by Nguyen, et al. in U.S. Pat. No. 6,249,073, that is not entirely determined via a thin sacrificial oxide, as was done (with difficulty) in previous clamped-clamped beam high frequency devices. Rather, as taught by Nguyen, et al., the capacitor gap 66 is determined by the height of spacers or dimples 68 set by a timed etch. The height of the dimples 68 is such that when a sufficiently large DC-bias V.sub.P is applied between the drive electrode 62 and the resonator beam 54, the whole structure comes down and rests upon the dimples 68, which are located at the flexural nodal points 56. The spacers 68 are formed either on the resonator beam 54 or on the substrate 59.
As taught by Nguyen, et al. in U.S. Pat. No. 6,249,073, the use of dimples to set the capacitor gap spacings 66 is intended to permit much thicker sacrificial oxide spacers to be used, thereby alleviating previous problems due to pinholes and non-uniformity in ultra-thin sacrificial layers used when transducer capacitor gap spacing is entirely determined by sacrificial surface micromachining for removing the thin sacrificial oxide. Also, the thicker sacrificial oxide is intended to be easier to remove than previous thinner ones, which is intended to decrease the required HF release etch time and lessen the probability that etching by-products remain in the gap 66 where they might interfere with resonator operation and Q.
FIGS. 6A, 6B and 6C illustrate one state of the art fabrication method for producing micromechanical resonators as taught by Nguyen, et al. in U.S. Pat. No. 6,249,073, wherein the device 52 is fabricated using a five-mask, polycrystalline silicon or “polysilicon,” surface-micromachining technology described by the process flow shown in U.S. Pat. No. 6,249,073. The fabrication sequence taught by Nguyen, et al. begins with isolation layers 70 and 72 formed via successive growth and deposition of 2 micron thermal oxide and 2000 Angstrom LPCVD Si.sub.3N.sub.4, respectively, over a <100> lightly-doped p-type starting silicon wafer 74. Next, 3000 Angstroms of LPCVD polysilicon is deposited at 585 degrees C. and phosphorous-doped via implantation, then patterned to form the ground planes 64 and interconnects. An LPCVD sacrificial oxide layer 78 is then deposited to a mathematically determined thickness, after which successive masking steps produce dimple and anchor openings 80, 82. The dimple openings 82 are defined via a reactive-ion etch which must be precisely controlled. Anchors openings 80 are simply wet-etched in a solution of buffered hydrofluoric acid (BHF).
Next, in FIG. 6B, structural polysilicon is deposited via LPCVD at 585 degrees C. and phosphorous dopants are introduced via ion-implantation to provide the flexural resonator beam 54. A 2000 Angstrom-thick oxide mask is then deposited via LPCVD at 900 degrees C., after which the wafers must be annealed for one hour at 1000 degrees C. in an effort to relieve stress and distribute dopants.
As illustrated in FIG. 6C, wet etching of the sacrificial layer is used for final release of the flexural resonator beam 54 to create the capacitor gap 66. Both the oxide mask and structural layer are patterned via SF.sub.6/O.sub.2 and Cl.sub.2-based RIE etches, respectively. The structures 54 and 58 are then released via a 5 minute etch in 48.8 wt. % HF. As taught by Nguyen, et al. in U.S. Pat. No. 6,249,073, this 5 minute release etch time is significantly shorter than that required for previous clamped-clamped beam resonators, which is about 1 hour, because they did not benefit from the dimple-activated gap spacings, as taught by Nguyen, et al. The previous clamped-clamped beam resonators require sacrificial oxide thicknesses on the order of hundreds of Angstroms. After structural release by wet etching of the sacrificial layer, aluminum is evaporated and patterned over polysilicon interconnects via lift-off to reduce series resistance.
Thus, state of the art methods of manufacturing using polycrystalline silicon produces resonators having intrinsic stresses, stress gradients, or both, that effect device performance. These state of the art methods of manufacturing using polycrystalline silicon also require removal of a sacrificial layer of material using wet etching techniques that complicate the manufacturing process and generally results in low yield due to difficulty in removing sacrificial material in the small gap between the resonator and the lower electrode. This method of manufacturing using polycrystalline silicon also results in stiction, which further lowers yields, and impurities present in polysilicon thin films degrade device performance and result in lower resonator Q.
Thus, an improved device and method of manufacturing are desirable.